Classifications

• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
  • H04N13/00—Stereoscopic video systems; Multi-view video systems; Details thereof
  • H04N13/20—Image signal generators
  • H04N13/275—Image signal generators from 3D object models, e.g. computer-generated stereoscopic image signals
• • G—PHYSICS
  • G01—MEASURING; TESTING
  • G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
  • G01B11/00—Measuring arrangements characterised by the use of optical techniques
  • G01B11/24—Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures
  • G01B11/25—Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures by projecting a pattern, e.g. one or more lines, moiré fringes on the object
  • G01B11/2504—Calibration devices
• • G—PHYSICS
  • G01—MEASURING; TESTING
  • G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
  • G01B11/00—Measuring arrangements characterised by the use of optical techniques
  • G01B11/24—Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures
  • G01B11/25—Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures by projecting a pattern, e.g. one or more lines, moiré fringes on the object
  • G01B11/2513—Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures by projecting a pattern, e.g. one or more lines, moiré fringes on the object with several lines being projected in more than one direction, e.g. grids, patterns
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T5/00—Image enhancement or restoration
  • G06T5/80—Geometric correction
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T7/00—Image analysis
  • G06T7/50—Depth or shape recovery
  • G06T7/521—Depth or shape recovery from laser ranging, e.g. using interferometry; from the projection of structured light

Definitions

• the present invention generally relates to the field of three-dimensional scanning of the surface geometry of an object, and, more particularly, to structured light stereoscopy.
• the shape of an object is scanned and digitized using a ranging sensor that measures the distance between the sensor and a set of points on the surface.
• a ranging sensor that measures the distance between the sensor and a set of points on the surface.
• Different principles have been developed for range sensors. Among them, interferometry, time-of-flight and triangulation-based principles are well-known principles that are each more or less appropriate depending on the requirements on accuracy, the stand-off distance between the sensor and the object, and the required depth of field.
• Some triangulation-based range sensors are generally adequate for close range measurements, such as inferior to a few meters.
• this type of apparatus at least two rays that converge to the same feature point on the object are obtained from two different viewpoints separated by a baseline distance. From the baseline and two ray directions, the relative position of the observed point can be recovered. The intersection of both rays is determined using the knowledge of one side length and two angles in the triangle, which actually is the principle of triangulation in stereovision. The challenge in stereovision is to efficiently identify which pixels correspond to each other in each image.
• one of the light detectors cameras
• a light projector that outputs a set of rays in known directions. In this case, it is possible to exploit the direction of the projected rays and each detected ray reflected on the object surface to solve the triangle. It is then possible to calculate the coordinates of each observed feature point relative to the basis of the triangle.
• the light source can be a coherent source (laser) or non-coherent source (e.g. white light) projecting a spot, a light plane or many other possible patterns of projection including a full-field pattern.
• a full-field pattern is a 2D pattern which can cover a portion or the whole of the projector's 2D field of illumination. In this case, a dense set of corresponding points can be matched in each image.
• Use of a light projector facilitates the detection of reflected points everywhere on the object surface so as to provide a dense set of measured surface points. However, the more complex the pattern will be, the greater the challenge will be to efficiently identify corresponding pixels and rays.
• the epipolar geometry can be computed from calibration parameters or after matching a set of points in two images.
• Another approach consists in rectifying the two images, which means all epipolar lines will be horizontal and aligned. Rectifying images is thus advantageous since no further calculation needs to be performed for identifying pixels on the epipolar lines.
• Image rectification can be applied by software or even by cautiously aligning the relative orientation of one or the two cameras (or projector). In this case, the approach is referred to as hardware alignment.
• a code initially aligned along a straight line (epipolar) in the projector image (hereafter referred to as slide image) will not be physically projected along a straight line after the lens and will thus not result in a good alignment along the epipolar line in the image of the camera.
• epipolar epipolar
• slide image a code initially aligned along a straight line
• distortion increases towards the side and corners of the images.
• One will either lose these points, compensate with larger bands for encoding the signal along the distorted epipolar lines (thus reducing resolution of measurement) or apply more complex calculations that will take away the initial goal of simplifying matching.
• a method for preparing a spatial coded slide image in which a pattern of the spatial coded slide image is aligned along epipolar lines at an output of a projector in a system for 3D measurement of a shape of an object, having the projector and a camera in full-field structured light comprising: obtaining a set of distortion vectors for projector coordinates of the projector, each the distortion vector representing a distortion from predicted coordinates caused by the projector; retrieving an ideal pattern image, wherein the ideal pattern image is an ideal image of the spatial coded pattern aligned on ideal epipolar lines; creating a real slide image by, for each real pixel coordinates of the real slide image, retrieving a current distortion vector from the set using the real pixel coordinates; removing distortion from the real pixel coordinates using the current distortion vector to obtain ideal pixel coordinates in the ideal pattern image; extracting a pixel value at the ideal pixel coordinates in the ideal pattern image; copying the pixel value at
• the step of creating a real slide image includes creating an electronic version of the real slide image and providing the electronic version to a programmable projector.
• the step of extracting a pixel value includes interpolating the pixel value.
• the pixel value is a level value.
• a method for facilitating matching of coded patterns between a projected image and a captured image in a system for 3D measurement of a shape of an object, having a projector and a camera in full-field structured light comprising: calibrating the projector and the camera for intrinsic and extrinsic parameters; preparing a spatial coded slide image in which a pattern of the spatial coded slide image is aligned along epipolar lines; projecting the spatial coded pattern on a scene object using the projector; observing the spatial coded pattern on the object using the camera to generate a camera image; processing the camera image to match codes with the projected image;
• the method further comprises undistorting and rectifying the camera image prior to the processing the camera image.
• a method for setting up a system for 3D measurement of a shape of an object having a projector with a fixed slide mask and a camera in full-field structured light, comprising: setting the lens aperture and focus; preparing a slide with a spatial coded slide image, mounting the slide rigidly with the projector lens and aligning a center of the slide with an optical axis of the lens; adjusting rotation around the optical axis of the lens and the translation of the projector along the optical axis of the lens so as to align the pattern code along the epipolar lines.
• FIG. 1 includes FIG. 1A, FIG. IB and FIG. 1C, wherein FIG. 1A is an illustration of a grid, FIG. IB is an illustration of the effect of barrel-type radial lens distortion and FIG. 1C is an illustration of the effect of pincushion radial lens distortion;
• FIG. 2 is a representation of the epipolar geometry
• FIG. 3 includes FIG. 3A and FIG. 3B, wherein FIG. 3A depicts a rear view of a rectified configuration and FIG. 3B depicts a top view of a rectified configuration;
• FIG. 4 is an illustration of the rectification process
• FIG. 5 illustrates distortion compensation applied to the slide image
• FIG. 6 illustrates the deformation of an epipolar line by a real lens
• FIG. 7 is a flowchart of an example method for producing the real pattern
• FIG. 8 includes FIG. 8A and FIG. 8B, wherein FIG. 8A is an illustration of a binary pattern and FIG. 8B is an illustration of the effect of thresholding after interpolation;
• FIG. 9 includes FIG. 9A and FIG. 9B, wherein FIG. 9A depicts a pre-distorted slide section along with its corresponding ideal section, FIG. 9B depicts the corresponding ideal section;
• FIG. 10 is a flowchart of an example method for adapting a fixed slide mask.
• the present invention allows aligning higher resolution code, even near the sides and corners of the image.
• the projector lens will distort the image built on the projector slide.
• the slide is the physical imager component that is located before the optics of the projector. It is either a transmitting or reflecting imager component.
• the pattern codes aligned along ideal epipolar lines on the slide will thus result in curved lines instead of straight lines once projected through the lens.
• the method therefore aligns the pattern codes with the actual epipolar lines after the lens instead of aligning the pattern codes on the hypothetical non-distorted straight lines on the projector slide.
• the distortion induced by the lens optics of the projector is first modeled and the distortion model is then applied to deform the coded patterns initially aligned along straight lines.
• the resulting coded patterns on the slide are thus pre-curved.
• the distortion of the projection lens then occurs as modeled and the coded patterns on the image captured by the camera are straightened.
• FIG. 1 shows an effect of radial lens distortion on a regular grid 101 shown in FIG. 1A.
• Radial distortion can lead to either barrel type distortion, shown at 102 in FIG. IB, or pincushion type distortion, shown at 103 in FIG. 1C.
• the effect is well-known. Straight lines are curved and the effect will be more important for short focal lengths.
• radial distortion is a very common type of distortion that is compensated for in machine vision and photogrammetry, other types of lens distortion can also be compensated for.
• One other such example lens distortion is tangential distortion.
• the projection model for both the camera and the projector is a pinhole with lens distortion mpensation.
• the pinhole model describes the relationship between a 3D point
• the intrinsic parameters, including distortion, as well as the geometric transformation between the projector and the camera can be calculated beforehand at calibration stage.
• the parameters describing this geometric transformation are referred to as the extrinsic parameters.
• a few methods are proposed in the art to obtain these parameters for a projector-camera combination. After obtaining the parameters, it is possible to calculate both the distorted and non-distorted pixels given one or the other.
• the projective geometry of two cameras describes the relationship between the positions of a point in one image with its corresponding point in the second image. Given a point in one image, its corresponding point lays along a straight line in the second image. This is illustrated in FIG. 2, where points O and O' are the projection centers of the devices and P is a point in 3D space.
• the set composed of P, O and O' defines an epipolar plane ⁇ that intersects both image planes ⁇ and ⁇ ' along lines 1 and . Lines 1 and are epipolar lines. So given a point p in the image 1 10, its corresponding point, p', in image 111 would be found along .
• the corresponding point to p' in image 110 would be found along line 1.
• These matrices can be obtained after calibration.
• the orientation of the epipolar lines is dictated by the vergence of the stereo arrangement. More precisely, when both image planes ⁇ and ⁇ ' are parallel, the epipolar lines will all be parallel as well. In the particular case of two parallel image planes ⁇ and ⁇ ' that are also parallel to the baseline and defined as the segment joining the two projection centers, the epipolar lines will be parallel to this baseline.
• image planes 113 and 114 can be adjusted in such a way that the epipolar lines will lay on the same lines in the images.
• the two image planes are then referred to as rectified.
• the epipolar plane is shown at 112.
• FIG. 3B a top view of the two parallel image planes is shown.
• FIG. 4 a pixel p in original image 110 is copied to its corresponding position, p rect , in the rectified image 1 13. The same principle would apply to the pair of images 11 1 and 114.
• a method is proposed to align the codes of the projector along the epipolar lines.
• the code then encodes a non-ambiguous position on a line compared with a position in the whole image.
• a system with spatial codes nearly aligned along the epipolar can be proposed to facilitate correspondence matching.
• the epipolar lines are not straight lines on the slide and they cannot be obtained simply by intersecting the epipolar plane with the image (slide) planes.
• the codes can be aligned along curved lines that will be straight (in a light plane) once outputted from the projector.
• the arrangement composed of the projector and camera is first calibrated for the intrinsic and extrinsic parameters. Then, considering an ideal image of the coded patterns on all ideal epipolar lines, typically in the rectified configuration, the image slide that is to be projected is the same image where the position of each pixel is corrected in the direction of 6(a,d). This is illustrated in FIG. 5.
• the ideal pinhole model is shown at 122.
• the output image after the pinhole is shown at 121 while the projected image is shown at 120.
• a spatial code provided on line 127 of the projected image 120 would be projected at line 123 on the output image 121.
• Line 127 would be chosen such that line 123 would be aligned on the epipolar line. However, to compensate for the distortion by the projector, rather than being provided on ideal line 127, the spatial code is rather aligned along one of the actual projected curves 124. This ensures that, after distortion, it is still projected at line 123 and therefore aligned on the epipolar line. For a given point, the vector between the ideal and distorted pixels is shown at 125. FIG. 6 illustrates the resulting effect with a real lens 126.
• Distortion vectors for the projector coordinates are first obtained. These can be determined, for example, using the projector model detailed above. As will be readily understood, other projector models could be used without departing from the invention with more or less radial and/or tangential terms and/or with other terms of distortion. Each distortion vector represents a distortion from predicted coordinates caused by the projector at the particular projector coordinates. After loading the ideal pattern image into memory 131, one will process, at 132, each pixel of the real slide by first removing distortion from the real pixel coordinates using the distortion vectors and obtaining the pixel coordinates in the ideal reference image 133.
• the optical axis will intersect the slide at its center. This intersection point defines the principal point of the slide.
• the pixel value from the ideal pattern image at these pixel coordinates will be obtained. This pixel value can be directly extracted (0-order interpolation) from the nearest pixel in the ideal image or it can be obtained using subpixel interpolation 134.
• the pixel value may be a level value representing color and/or intensity.
• the pixel value is finally copied to the current pixel in the real slide image 135. This process is repeated for all pixels of the real slide image 136.
• Some coded patterns may be binary images to increase the signal-to-noise ratio or to get increased precision when 3D positions are calculated from points located at the image edges.
• the process that has just been described will work well to compensate lens distortion, the resulting image is obtained after subpixel interpolation, which will introduce gray level pixels even if the ideal pattern is binary. Imposing a binary value by thresholding will deform the shape of the edge in the projected image.
• FIG. 8A an example of an ideal spatial code is shown at 140.
• FIG. 8B the potential effect of thresholding is shown.
• Fig. 9A illustrates a section of the resulting slide at 150.
• the expected projected pattern that will be "undistorted" by the optics is shown at 151 in FIG. 9B.
• Two radial terms (ki and k 2 ) were used to generate the slide section shown at 150 in FIG. 9A.
• the modeled lens is a fixed focal length Fujinon 9 mm, model HF9HA-1B f/1.4 exploited at f/2 and focalized at a distance of 350 mm.
• More continuous horizontal edges can also be obtained after dividing each rectangle into several narrower subrectangles with the same height and applying the same procedure to each of these subrectangles. This is especially interesting for a fixed slide mask where the resolution is usually higher than most programmable projectors. Other embodiments of the method are also possible to reach the same goal.
• the pattern can be reconfigured at runtime.
• the pattern code can be adapted based on the calibration parameters.
• FIG. 10 An example method 160 for setting up a system for 3D measurement is shown in Fig. 10.
• the distortion parameters of the lens are calibrated beforehand at 162, after the aperture and focus have been adjusted at 161.
• the fixed slide mask is created based on these parameters and using the example procedure detailed in Fig. 7 and represented in Fig. 10 at 163. This procedure was carried out to produce the image shown at 150 in FIG. 9A with the Fujinon lens described above.
• the mask is mounted with the lens, and the center of distortion is precisely aligned at 164.
• a homography, H is a one-to-one projective mapping between 2D coordinates of the mask and the camera image. It is thus possible to align the principal point of the slide with the optical axis of the lens. Actually, when the optical axis intersects with the principal point of the slide, the following expression is minimized:
• a p is a point on the projector slide after removing distortion using the projector distortion model
• a c is a point in the camera image after removing the distortion using the camera distortion model
• Ha c is the point a c mapped to the undistorted projector slide
• ⁇ is a set of matched points between the projector slide and the camera image.

Landscapes

• Engineering & Computer Science (AREA)
• Physics & Mathematics (AREA)
• General Physics & Mathematics (AREA)
• Computer Vision & Pattern Recognition (AREA)
• Theoretical Computer Science (AREA)
• Multimedia (AREA)
• Signal Processing (AREA)
• Optics & Photonics (AREA)
• Length Measuring Devices By Optical Means (AREA)
• Image Processing (AREA)
• Measurement Of Optical Distance (AREA)

Applications Claiming Priority (2)

Publications (3)

Family

ID=48469222

Family Applications (1)

Country Status (5)

Families Citing this family (20)

Family Cites Families (16)

• 2012
  • 2012-11-02 CN CN201280049264.5A patent/CN103988048B/zh active Active
  • 2012-11-02 EP EP12852417.0A patent/EP2751521B1/de active Active
  • 2012-11-02 JP JP2014535225A patent/JP5635218B1/ja active Active
  • 2012-11-02 US US13/699,773 patent/US8836766B1/en active Active
  • 2012-11-02 WO PCT/IB2012/056112 patent/WO2013076605A1/en active Application Filing

Non-Patent Citations (2)

Also Published As

Similar Documents

Legal Events